It is true that if $f$ is a real valued function, and if the iterated limits $\lim_{x\rightarrow x_0}[\lim_{y\rightarrow y_0}f(x,y)]$ and $\lim_{y\rightarrow y_0}[\lim_{x\rightarrow x_0}f(x,y)]$ exist and are equal, we cannot say anything about the double limit $\lim_{(x,y)\rightarrow (x_0,y_0)} f(x,y)$ .

But look at this proof. Let $\lim_{y\rightarrow y_0}f(x,y)=l_{y_0}(x)$, $\lim_{x\rightarrow x_0}f(x,y) = l_{x_0}(y)$, $\lim_{y\rightarrow y_0}[\lim_{x\rightarrow x_0}f(x,y)] = l_{x_0 y_0}$

Let $\epsilon$> 0. $\exists \ \delta _1 >0$ st whenever $|x-x_0|< \delta_1$ we have $|f(x,y)-l_{x_0}(y)|<\epsilon$ and $\exists$ $\delta _2 >0$ st whenever $|y-y_0|< \delta_2$ we have $|l_{x_0 y_0}-l_{x_0}(y)|<\epsilon$.

Then we have $|f(x,y)-l_{x_0 y_0}|\leq |f(x,y)-l_{x_0}(y)|+ |l_{x_0 y_0}-l_{x_0}(y)|<2\epsilon$ for $|x-x_0|<\delta_3$ and $|y-y_0|<\delta_3$ where $\delta_3= min(\delta_1, \delta_2)$

We can proceed similarly with $\lim_{x\rightarrow x_0}[\lim_{y\rightarrow y_0}f(x,y)]$ and get the same result (although it is not necessary). Thus ,we have shown the double limit is indeed the same as the iterated limit. Where did I go wrong in the proof?

  • 1
    $\begingroup$ Your definition of $l_{y_0}$ is unclear. Where did the $x$ go? The expression $\lim_{y\rightarrow y_0} f(x,y)$ assumes the limit exists for all $x$, and in general the limit would depend on $x$. So it would make more sense to say $\lim_{y\rightarrow y_0} f(x,y) = l_{y_0, x}$ (assuming the limit exists). Example: If $f(x,y)=xy$ then $\lim_{y\rightarrow y_0} f(x,y)= xy_0$. $\endgroup$
    – Michael
    Oct 27, 2020 at 13:26
  • $\begingroup$ @Michael I have made the necessary edits. Thanks for pointing that out. $\endgroup$
    – suraj
    Oct 27, 2020 at 13:41
  • 1
    $\begingroup$ As mentioned in an answer below, you really have $\delta_1(y)$ and $\delta_2(x)$. You would need to again rewrite the proof. Also, in your paragraph "Then we have ...", it is not clear what assumptions you make on $x$ and $y$. $\endgroup$
    – Michael
    Oct 27, 2020 at 14:13
  • $\begingroup$ @Michael I think I get it now. Thanks. $\endgroup$
    – suraj
    Oct 27, 2020 at 14:19

1 Answer 1


In your proof, $\delta_1$ depends on $y$. So, it may happen that $\delta_1$ shrinks to $0$, as $y\to y_0$. Consider the function $f\colon \mathbb{R}^2\to \mathbb{R}$,

$f(x,y)=\frac{x^2y^2}{x^2y^2+(x-y)^2}$ for $x,y\neq 0$ and $f(0,0):=0$.

Let $(x_0,y_0):=(0,0)$. Now, it is easy to see, that (in your notation) $l_{x_0}=l_{y_0}=l_{x_0 y_0}=0$, but $f(\delta,\delta)=1$ for any $\delta>0$, so that $\lim_{(x,y)\to(x_0,y_0)}f(x,y)$ does not exist (note that $f(0,0)=0$ by definition). Now, why does your proof don't work in this example? Well, for $0<\varepsilon<1$, no matter how small you choose $\delta_1>0$, we have for $x$ such that $|x-x_0|<\delta_1$ and $y=x$


So in this example, if $y$ is of the same magnitude as $x$, then you cannot ensure $|f(x,y)-l_{x_0}(y)|<\varepsilon$ for $|x-x_0|<\delta_1$. You can only ensure it if you fix $y$, i.e., $\delta_1=\delta_1(y)$ depends on $y$ and $\delta_1(y)\to 0$ as $y\to y_0$ so that $\delta_3(y)=\min\{\delta_1(y), \delta_2\}\to 0$ as $y\to y_0$, contradicting the requirement that $\delta_3$ has to be strictly positive.

  • $\begingroup$ But when y is in a neighborhood of y_0, \delta_1 must be non zero $\endgroup$
    – suraj
    Oct 27, 2020 at 13:58
  • $\begingroup$ @suraj : You are stating that $\delta_1(y)>0$ for all $y$. Fine. But so what? $\endgroup$
    – Michael
    Oct 27, 2020 at 14:14
  • $\begingroup$ @suraj: I editted my answer and gave an example. I hope that clarifies more why your proof does not work. $\endgroup$
    – ym94
    Oct 27, 2020 at 14:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.